Grenfell Norman Patey

Organic ice-nucleating substances (INSs) are relevant to atmospheric, biological, and industrial systems. Fatty acid and fatty alcohol monolayers have been used as model surfaces to investigate ice nucleation by organic materials. Laboratory studies show that fatty acid monolayers are inefficient ice nucleators, whereas fatty alcohol monolayers with similar alkyl-chain lengths can nucleate ice at temperatures up to approximately 20 °C warmer. Here, we use molecular dynamics simulations to investigate the molecular origin of this difference for a model fatty acid monolayer. Our results show that a major factor inhibiting ice nucleation by fatty acid monolayers is the presence of partially charged carbonyl oxygen atoms. The electrostatic interactions of these atoms with interfacial water molecules is consistent with strong carbonyl-water hydrogen bonding, and disrupts the formation of ordered, ice-like bilayers that are critical for nucleation on fatty alcohol monolayers. This conclusion is supported by simulations in which removing the partial charges on the carbonyl oxygen atoms restores ice nucleation. In contrast, differences in lattice mismatch and headgroup structure play a lesser role. These findings provide molecular-level insight into how interfacial functional groups control the ice nucleating ability of organic surfaces.

Molecular dynamics (MD) simulations are employed to examine ice nucleation by amphiphilic alcohol monolayers formed by two alkyl hydroxy esters and two bromo alcohols. These monolayers expose hydroxylated surfaces to water. Despite their poor two-dimensional (2D) lattice matches to hexagonal ice, these surfaces nucleate ice at surprisingly warm temperatures. In our simulations, monolayers with experimental lattice parameters did not nucleate hexagonal ice, but one alkyl hydroxy ester and one bromo alcohol surface nucleated cubic ice via its (001) face. However, cubic ice nucleation required particular surface conformations that were not available for the other two monolayers. We also consider the possibility that the lattice parameters measured at 5 °C do not give an accurate representation of the lattices at the lower (-5 °C to -9 °C) ice nucleating temperatures (INTs). We use measurements only available for a single alcohol monolayer (C₂₁H₄₃OH) to estimate changes in the lattice parameters of the alkyl hydroxy ester and bromo alcohol monolayers at their INTs. When cooled from 5 °C to their INTs, the monolayer lattice parameter contractions vary from 3.7% to 5.6%, substantially improving the 2D lattice matches to hexagonal ice. MD simulations performed by varying the relevant lattice parameter, a, showed that hexagonal ice nucleation occurs if a is in the range 5.0 Å - 5.6 Å. For the alkyl hydroxy ester monolayers, the lattice parameters estimated at their INTs (5.44 Å and 5.48 Å) lie well within the hexagonal ice-nucleating region. For the bromo alcohol monolayers, the estimates (5.67 Å and 5.61 Å) are borderline. Overall, our analysis suggests that the temperature dependence of lattice parameters is a significant factor contributing to ice nucleation by alkyl hydroxy ester and bromo alcohol monolayers, a possibility not previously considered.

Short-term exposure to ammonium sulfate enhances ice nucleation by submicrometer α-alumina but not by alcohol monolayers or microplastics, supporting ammonium sulfate assays for identifying mineral ice nucleators.

Organic ice nucleating substances (INSs) are thought to play an essential role in cloud formation and, hence, precipitation and climate. Organic INSs are an important but poorly understood class of INSs in the atmosphere. To study organic INSs with exposed hydroxylated surfaces, researchers have previously used fatty alcohol monolayers as model systems. For alcohol monolayers, ice nucleation temperatures increase with increasing alkyl chain length and show a high-low oscillation following the number (odd-even) of carbon atoms in the alkyl chains. We employ atomistic models, together with molecular dynamics simulations, to investigate ice nucleation by C₂₀H₄₁OH, C₃₀H₆₁OH, and C₃₁H₆₃OH monolayers. As expected, we find that ice nucleation by alcohol monolayers depends on the lattice match to ice, and a poorer lattice match can at least partially account for the reduced ice nucleation ability of C₂₀H₄₁OH monolayers compared to monolayers of the longer chain alcohols. More interestingly, our simulations identify a limited range of alcohol configurations that readily nucleate ice via the basal plane. For configurations outside this range, ice nucleation did not occur on the time scale of our simulations (i.e., 5000 ns). The configurational feature that crucially influences ice nucleation is the angle between the alcohol C-O bond and the interfacial plane. C-O bonds directed sharply toward or away from the water phase strongly inhibit ice nucleation. In contrast, ice nucleation is easily observed for a relatively narrow band of C-O bond orientations centered about the surface plane. For comparable surface configurations, the ice nucleating abilities of C₃₀H₆₁OH and C₃₁H₆₃OH monolayers are practically identical, but the existence of a narrow band of ice-compatible surface configurations can perhaps explain why odd-chain alcohol monolayers are better INSs than even-chain alcohol monolayers. Earlier simulations have shown that for alcohols differing by a single carbon atom, the odd-chain monolayer is less rigid than the even-chain monolayer. This suggests the possibility that for odd-chain alcohol monolayers, the orientation of the C-O bonds can more easily adjust into the ice-compatible range than their even-chain counterparts, accounting for their enhanced ice nucleating ability.

Heterogeneous ice nucleation (HIN) has applications in climate science, nanotechnology, and cryopreservation. Ice nucleation on the earth’s surface or in the atmosphere usually occurs heterogeneously involving foreign substrates, known as ice nucleating particles (INPs). Experiments identify good INPs but lack sufficient microscopic resolution to answer the basic question: What makes a good INP? We employ molecular dynamics (MD) simulations in combination with machine learning (ML) to address this question. Often, the large amount of computational cost required to cross the nucleation barrier and observe HIN in MD simulations is a practical limitation. We use information obtained from short MD simulations of atomistic surface and water models to predict the likelihood of HIN. We consider 153 atomistic substrates with some surfaces differing in elemental composition and others only in terms of lattice parameters, surface morphology, or surface charges. A range of water features near the surface (local) are extracted from short MD simulations over a time interval (≤300 ns) where ice nucleation has not initiated. Three ML classification models, Random Forest (RF), support vector machine, and Gaussian process classification are considered, and the accuracies achieved by all three approaches lie within their statistical uncertainties. Including local water features is essential for accurate prediction. The accuracy of our best RF classification model obtained including both surface and local water features is 0.89 ± 0.05. A similar accuracy can be achieved including only local water features, suggesting that the important surface properties are largely captured by the local water features. Some important features identified by ML analysis are local icelike structures, water density and polarization profiles perpendicular to the surface, and the two-dimensional lattice match to ice. We expect that this work, with its strong focus on realistic surface models, will serve as a guide to the identification or design of substrates that can promote or discourage ice nucleation.

In mixed-phase or ice clouds, ice can be formed through heterogeneous nucleation. A major type of ice-nucleating particle (INP) in the atmosphere are mineral dust particles. For mixed-phase clouds, the pH of water droplets can vary widely and influence ice nucleation by altering the surface of some INPs, including mineral dust. Kaolinite is a commonly occurring clay mineral, and laboratory experiments, as well as molecular dynamics (MD) simulations, have demonstrated its ice-nucleating efficiency at neutral pH. We examine the influence of pH on the ice-nucleating efficiency of kaolinite, in the immersion freezing mode, through both droplet freezing experiments and MD simulations. Droplet freezing experiments using KGa-1b kaolinite samples are reported under both acidic (HNO3 solutions) and basic (NaOH solutions) conditions, covering the measured pH range 0.18-13.26. These experiments show that the ice-nucleating efficiency of kaolinite is not significantly influenced by the presence of acid but is reduced in extremely basic conditions. We report MD simulations aimed at gaining a microscopic understanding of the pH dependence of ice nucleation by kaolinite. The Al(001), Si(001), and three edge surfaces of kaolinite are considered, but ice nucleation was observed only for the Al(001) surface. The hydroxy groups exposed on the Al(001) surface can be deprotonated in a basic solution or dual-protonated in an acidic solution, which can influence ice nucleation efficiency. The protonation state of the Al(001) surface for a particular pH can be estimated using previously measured pKa values. We find that the monoprotonated Al(001) surface expected to be stable at near-neutral pH is the most effective ice-nucleating surface. In MD simulations, the ice nucleation efficiency persists for dual-protonation but decreases significantly with increasing deprotonation, qualitatively consistent with the experimental observations. Taken together, our experimental and MD results for a wide range of pH values support the suggestion that the Al(001) surface may be important for ice nucleation by kaolinite. Additionally, the deprotonation of hydroxy groups on INP surfaces can have a significant effect on their ice-nucleating ability.

Convolutional neural networks are constructed and validated for the crystal structure classification of simple binary salts such as the alkali halides. The inputs of the neural network classifiers are the local bond orientational order parameters of Steinhardt, Nelson, and Ronchetti [Phys. Rev. B 28, 784 (1983)], which are derived solely from the relative positions of atoms surrounding a central reference atom. This choice of input gives classifiers that are invariant to density, increasing their transferability. The neural networks are trained and validated on millions of data points generated from a large set of molecular dynamics (MD) simulations of model alkali halides in nine bulk phases (liquid, rock salt, wurtzite, CsCl, 5-5, sphalerite, NiAs, AntiNiAs, and β-BeO) across a range of temperatures. One-dimensional time convolution is employed to filter out short-lived structural fluctuations. The trained neural networks perform extremely well, with accuracy up to 99.99% on a balanced validation dataset constructed from millions of labeled bulk phase structures. A typical analysis using the neural networks, including neighbor list generation, order parameter calculation, and class inference, is computationally inexpensive compared to MD simulations. As a demonstration of their accuracy and utility, the neural network classifiers are employed to follow the nucleation and crystal growth of two model alkali halide systems, crystallizing into distinct structures from the melt. We further demonstrate the classifiers by implementing them in automated MD melting point calculations. Melting points for model alkali halides using the most commonly employed rigid-ion interaction potentials are reported and discussed.